[dpdk-dev] [PATCH v3] distributor_app: new sample app

[Neil Horman]

On Tue, Sep 30, 2014 at 11:39:37AM +0100, reshmapa wrote:
> From: Reshma Pattan <reshma.pattan at intel.com>
> 
> A new sample app that shows the usage of the distributor library. This
> app works as follows:
> 
> * An RX thread runs which pulls packets from each ethernet port in turn
>   and passes those packets to worker using a distributor component.
> * The workers take the packets in turn, and determine the output port
>   for those packets using basic l2forwarding doing an xor on the source
>   port id.
> * The RX thread takes the returned packets from the workers and enqueue
>   those packets into an rte_ring structure.
> * A TX thread pulls the packets off the rte_ring structure and then
>   sends each packet out the output port specified previously by the worker
> * Command-line option support provided only for portmask.
> 
> Signed-off-by: Bruce Richardson <bruce.richardson at intel.com>
> Signed-off-by: Reshma Pattan    <reshma.pattan at intel.com>
> ---
>  examples/Makefile                 |    1 +
>  examples/distributor_app/Makefile |   57 ++++
>  examples/distributor_app/main.c   |  600 +++++++++++++++++++++++++++++++++++++
>  examples/distributor_app/main.h   |   46 +++
>  4 files changed, 704 insertions(+), 0 deletions(-)
>  create mode 100644 examples/distributor_app/Makefile
>  create mode 100644 examples/distributor_app/main.c
>  create mode 100644 examples/distributor_app/main.h
> 
> diff --git a/examples/Makefile b/examples/Makefile
> index 6245f83..2ba82b0 100644
> --- a/examples/Makefile
> +++ b/examples/Makefile
> @@ -66,5 +66,6 @@ DIRS-y += vhost
>  DIRS-$(CONFIG_RTE_LIBRTE_XEN_DOM0) += vhost_xen
>  DIRS-y += vmdq
>  DIRS-y += vmdq_dcb
> +DIRS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) += distributor_app
>  
>  include $(RTE_SDK)/mk/rte.extsubdir.mk
> diff --git a/examples/distributor_app/Makefile b/examples/distributor_app/Makefile
> new file mode 100644
> index 0000000..6a5bada
> --- /dev/null
> +++ b/examples/distributor_app/Makefile
> @@ -0,0 +1,57 @@
> +#   BSD LICENSE
> +#
> +#   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
> +#   All rights reserved.
> +#
> +#   Redistribution and use in source and binary forms, with or without
> +#   modification, are permitted provided that the following conditions
> +#   are met:
> +#
> +#     * Redistributions of source code must retain the above copyright
> +#       notice, this list of conditions and the following disclaimer.
> +#     * Redistributions in binary form must reproduce the above copyright
> +#       notice, this list of conditions and the following disclaimer in
> +#       the documentation and/or other materials provided with the
> +#       distribution.
> +#     * Neither the name of Intel Corporation nor the names of its
> +#       contributors may be used to endorse or promote products derived
> +#       from this software without specific prior written permission.
> +#
> +#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
> +#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> +#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
> +#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
> +#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
> +#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
> +#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
> +#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
> +#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> +#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
> +#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> +
> +ifeq ($(RTE_SDK),)
> +$(error "Please define RTE_SDK environment variable")
> +endif
> +
> +# Default target, can be overriden by command line or environment
> +RTE_TARGET ?= x86_64-native-linuxapp-gcc
> +
> +include $(RTE_SDK)/mk/rte.vars.mk
> +
> +# binary name
> +APP = distributor_app
> +
> +# all source are stored in SRCS-y
> +SRCS-y := main.c
> +
> +CFLAGS += $(WERROR_FLAGS)
> +
> +# workaround for a gcc bug with noreturn attribute
> +# http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
> +ifeq ($(CONFIG_RTE_TOOLCHAIN_GCC),y)
> +CFLAGS_main.o += -Wno-return-type
> +endif
> +
> +EXTRA_CFLAGS += -O3 -Wfatal-errors
> +
> +include $(RTE_SDK)/mk/rte.extapp.mk
> diff --git a/examples/distributor_app/main.c b/examples/distributor_app/main.c
> new file mode 100644
> index 0000000..f555d93
> --- /dev/null
> +++ b/examples/distributor_app/main.c
> @@ -0,0 +1,600 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
> + *   All rights reserved.
> + *
> + *   Redistribution and use in source and binary forms, with or without
> + *   modification, are permitted provided that the following conditions
> + *   are met:
> + *
> + *     * Redistributions of source code must retain the above copyright
> + *       notice, this list of conditions and the following disclaimer.
> + *     * Redistributions in binary form must reproduce the above copyright
> + *       notice, this list of conditions and the following disclaimer in
> + *       the documentation and/or other materials provided with the
> + *       distribution.
> + *     * Neither the name of Intel Corporation nor the names of its
> + *       contributors may be used to endorse or promote products derived
> + *       from this software without specific prior written permission.
> + *
> + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
> + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
> + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
> + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
> + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
> + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
> + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
> + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
> + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> + */
> +
> +#include <stdint.h>
> +#include <inttypes.h>
> +#include <unistd.h>
> +#include <signal.h>
> +#include <getopt.h>
> +
> +#include <rte_eal.h>
> +#include <rte_ethdev.h>
> +#include <rte_cycles.h>
> +#include <rte_malloc.h>
> +#include <rte_debug.h>
> +#include <rte_distributor.h>
> +
> +#include "main.h"
> +
> +#define RX_RING_SIZE 256
> +#define RX_FREE_THRESH 32
> +#define RX_PTHRESH 8
> +#define RX_HTHRESH 8
> +#define RX_WTHRESH 0
> +
> +#define TX_RING_SIZE 512
> +#define TX_FREE_THRESH 32
> +#define TX_PTHRESH 32
> +#define TX_HTHRESH 0
> +#define TX_WTHRESH 0
> +#define TX_RSBIT_THRESH 32
> +#define TX_Q_FLAGS (ETH_TXQ_FLAGS_NOMULTSEGS | ETH_TXQ_FLAGS_NOVLANOFFL |\
> +	ETH_TXQ_FLAGS_NOXSUMSCTP | ETH_TXQ_FLAGS_NOXSUMUDP | \
> +	ETH_TXQ_FLAGS_NOXSUMTCP)
> +
> +#define NUM_MBUFS ((64*1024)-1)
> +#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
> +#define MBUF_CACHE_SIZE 250
> +#define BURST_SIZE 32
> +#define RTE_RING_SZ 1024
> +
> +/* uncommnet below line to enable debug logs */
> +/* #define DEBUG */
> +
> +#ifdef DEBUG
> +#define LOG_LEVEL RTE_LOG_DEBUG
> +#define LOG_DEBUG(log_type, fmt, args...) do {	\
> +	RTE_LOG(DEBUG, log_type, fmt, ##args)		\
> +} while (0)
> +#else
> +#define LOG_LEVEL RTE_LOG_INFO
> +#define LOG_DEBUG(log_type, fmt, args...) do {} while (0)
> +#endif
> +
> +#define RTE_LOGTYPE_DISTRAPP RTE_LOGTYPE_USER1
> +
> +/* mask of enabled ports */
> +static uint32_t enabled_port_mask = 0;
> +
> +static volatile struct app_stats {
> +	struct {
> +		uint64_t rx_pkts;
> +		uint64_t returned_pkts;
> +		uint64_t enqueued_pkts;
> +	} rx __rte_cache_aligned;
> +
> +	struct {
> +		uint64_t dequeue_pkts;
> +		uint64_t tx_pkts;
> +	} tx __rte_cache_aligned;
> +} app_stats;
> +
> +static const struct rte_eth_conf port_conf_default = {
> +	.rxmode = {
> +		.mq_mode = ETH_MQ_RX_RSS,
> +		.max_rx_pkt_len = ETHER_MAX_LEN,
> +		.split_hdr_size = 0,
> +		.header_split   = 0, /**< Header Split disabled */
> +		.hw_ip_checksum = 0, /**< IP checksum offload enabled */
> +		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
> +		.jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
> +		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
> +	},
> +	.txmode = {
> +		.mq_mode = ETH_MQ_TX_NONE,
> +	},
> +	.lpbk_mode = 0,
> +	.rx_adv_conf = {
> +			.rss_conf = {
> +				.rss_hf = ETH_RSS_IPV4 | ETH_RSS_IPV6 |
> +					ETH_RSS_IPV4_TCP | ETH_RSS_IPV4_UDP |
> +					ETH_RSS_IPV6_TCP | ETH_RSS_IPV6_UDP,
> +			}
> +	},
> +};
> +
> +static const struct rte_eth_rxconf rx_conf_default = {
> +	.rx_thresh = {
> +		.pthresh = RX_PTHRESH,
> +		.hthresh = RX_HTHRESH,
> +		.wthresh = RX_WTHRESH,
> +	},
> +	.rx_free_thresh = RX_FREE_THRESH,
> +	.rx_drop_en = 0,
> +};
> +
> +static const struct rte_eth_txconf tx_conf_default = {
> +	.tx_thresh = {
> +		.pthresh = TX_PTHRESH,
> +		.hthresh = TX_HTHRESH,
> +		.wthresh = TX_WTHRESH,
> +	},
> +	.tx_free_thresh = TX_FREE_THRESH,
> +	.tx_rs_thresh = TX_RSBIT_THRESH,
> +	.txq_flags = TX_Q_FLAGS
> +
> +};
> +
> +struct output_buffer {
> +	unsigned count;
> +	struct rte_mbuf *mbufs[BURST_SIZE];
> +};
> +
> +/*
> + * Initialises a given port using global settings and with the rx buffers
> + * coming from the mbuf_pool passed as parameter
> + */
> +static inline int
> +port_init(uint8_t port, struct rte_mempool *mbuf_pool)
> +{
> +	struct rte_eth_conf port_conf = port_conf_default;
> +	const uint16_t rxRings = 1, txRings = rte_lcore_count() - 1;
> +	int retval;
> +	uint16_t q;
> +
> +	if (port >= rte_eth_dev_count())
> +		return -1;
> +
> +	retval = rte_eth_dev_configure(port, rxRings, txRings, &port_conf);
> +	if (retval != 0)
> +		return retval;
> +
> +	for (q = 0; q < rxRings; q++) {
> +		retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
> +						rte_eth_dev_socket_id(port),
> +						&rx_conf_default, mbuf_pool);
> +		if (retval < 0)
> +			return retval;
> +	}
> +
> +	for (q = 0; q < txRings; q++) {
> +		retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
> +						rte_eth_dev_socket_id(port),
> +						&tx_conf_default);
> +		if (retval < 0)
> +			return retval;
> +	}
> +
> +	retval  = rte_eth_dev_start(port);
> +	if (retval < 0)
> +		return retval;
> +
> +	struct rte_eth_link link;
> +	rte_eth_link_get_nowait(port, &link);
> +	if (!link.link_status) {
> +		sleep(1);
> +		rte_eth_link_get_nowait(port, &link);
> +	}
> +
> +	if (!link.link_status) {
> +		printf("Link down on port %"PRIu8"\n", port);
> +		return 0;
> +	}
> +
> +	struct ether_addr addr;
> +	rte_eth_macaddr_get(port, &addr);
> +	printf("Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
> +			" %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
> +			(unsigned)port,
> +			addr.addr_bytes[0], addr.addr_bytes[1],
> +			addr.addr_bytes[2], addr.addr_bytes[3],
> +			addr.addr_bytes[4], addr.addr_bytes[5]);
> +
> +	rte_eth_promiscuous_enable(port);
> +
> +	return 0;
> +}
> +
> +struct lcore_params {
> +	unsigned worker_id;
> +	struct rte_distributor *d;
> +	struct rte_ring *r;
> +};
> +
> +static __attribute__((noreturn)) void
> +lcore_rx(struct lcore_params *p)
> +{
> +	struct rte_distributor *d = p->d;
> +	struct rte_ring *r = p->r;
> +	const uint8_t nb_ports = rte_eth_dev_count();
> +	const int socket_id = rte_socket_id();
> +	uint8_t port;
> +
> +	for (port = 0; port < nb_ports; port++) {
> +		/* skip ports that are not enabled */
> +		if ((enabled_port_mask & (1 << port)) == 0)
> +			continue;
> +
> +		if (rte_eth_dev_socket_id(port) > 0 &&
> +				rte_eth_dev_socket_id(port) != socket_id)
> +			printf("WARNING, port %u is on remote NUMA node to "
> +					"RX thread.\n\tPerformance will not "
> +					"be optimal.\n", port);
> +	}
> +
> +	printf("\nCore %u doing packet RX.\n", rte_lcore_id());
> +	port = 0;
> +	for (;;) {
> +		/* skip ports that are not enabled */
> +		if ((enabled_port_mask & (1 << port)) == 0) {
> +			if (++port == nb_ports)
> +				port = 0;
> +			continue;
> +		}
> +		struct rte_mbuf *bufs[BURST_SIZE*2];
> +		const uint16_t nb_rx = rte_eth_rx_burst(port, 0, bufs,
> +				BURST_SIZE);
> +		app_stats.rx.rx_pkts += nb_rx;
> +
> +		rte_distributor_process(d, bufs, nb_rx);
> +		const uint16_t nb_ret = rte_distributor_returned_pkts(d,
> +				bufs, BURST_SIZE*2);
> +		app_stats.rx.returned_pkts += nb_ret;
> +		if (unlikely(nb_ret == 0))
> +			continue;
> +
> +		uint16_t sent = rte_ring_enqueue_burst(r, (void *)bufs, nb_ret);
> +		app_stats.rx.enqueued_pkts += sent;
> +		if (unlikely(sent < nb_ret)) {
> +			LOG_DEBUG(DISTRAPP, "%s:Packet loss due to full ring\n", __func__);
> +			while (sent < nb_ret)
> +				rte_pktmbuf_free(bufs[sent++]);
> +		}
> +		if (++port == nb_ports)
> +			port = 0;
> +	}
> +}
> +
> +static inline void
> +flush_one_port(struct output_buffer *outbuf, uint8_t outp)
> +{
> +	unsigned nb_tx = rte_eth_tx_burst(outp, 0, outbuf->mbufs,
> +			outbuf->count);
> +	app_stats.tx.tx_pkts += nb_tx;
> +
> +	if (unlikely(nb_tx < outbuf->count)) {
> +		LOG_DEBUG(DISTRAPP, "%s:Packet loss with tx_burst\n", __func__);
> +		do {
> +			rte_pktmbuf_free(outbuf->mbufs[nb_tx]);
> +		} while (++nb_tx < outbuf->count);
> +	}
> +	outbuf->count = 0;
> +}
> +
> +static inline void
> +flush_all_ports(struct output_buffer *tx_buffers, uint8_t nb_ports)
> +{
> +	uint8_t outp;
> +	for (outp = 0; outp < nb_ports; outp++) {
> +		/* skip ports that are not enabled */
> +		if ((enabled_port_mask & (1 << outp)) == 0)
> +			continue;
> +
> +		if (tx_buffers[outp].count == 0)
> +			continue;
> +
> +		flush_one_port(&tx_buffers[outp], outp);
> +	}
> +}
> +
> +static __attribute__((noreturn)) void
> +lcore_tx(struct rte_ring *in_r)
> +{
> +	static struct output_buffer tx_buffers[RTE_MAX_ETHPORTS];
> +	const uint8_t nb_ports = rte_eth_dev_count();
> +	const int socket_id = rte_socket_id();
> +	uint8_t port;
> +
> +	for (port = 0; port < nb_ports; port++) {
> +		/* skip ports that are not enabled */
> +		if ((enabled_port_mask & (1 << port)) == 0)
> +			continue;
> +
> +		if (rte_eth_dev_socket_id(port) > 0 &&
> +				rte_eth_dev_socket_id(port) != socket_id)
> +			printf("WARNING, port %u is on remote NUMA node to "
> +					"TX thread.\n\tPerformance will not "
> +					"be optimal.\n", port);
> +	}
> +
> +	printf("\nCore %u doing packet TX.\n", rte_lcore_id());
> +	for (;;) {
> +		for (port = 0; port < nb_ports; port++) {
> +			/* skip ports that are not enabled */
> +			if ((enabled_port_mask & (1 << port)) == 0)
> +				continue;
> +
> +			struct rte_mbuf *bufs[BURST_SIZE];
> +			const uint16_t nb_rx = rte_ring_dequeue_burst(in_r,
> +					(void *)bufs, BURST_SIZE);
> +			app_stats.tx.dequeue_pkts += nb_rx;
> +
> +			/* if we get no traffic, flush anything we have */
> +			if (unlikely(nb_rx == 0)) {
> +				flush_all_ports(tx_buffers, nb_ports);
> +				continue;
> +			}
> +
> +			/* for traffic we receive, queue it up for transmit */
> +			uint16_t i;
> +			_mm_prefetch(bufs[0], 0);
> +			_mm_prefetch(bufs[1], 0);
> +			_mm_prefetch(bufs[2], 0);
> +			for (i = 0; i < nb_rx; i++) {
> +				struct output_buffer *outbuf;
> +				uint8_t outp;
> +				_mm_prefetch(bufs[i + 3], 0);
> +				/* workers should update in_port to hold the
> +				 * output port value */
> +				outp = bufs[i]->port;
> +				/* skip ports that are not enabled */
> +				if ((enabled_port_mask & (1 << outp)) == 0)
> +					continue;
> +
> +				outbuf = &tx_buffers[outp];
> +				outbuf->mbufs[outbuf->count++] = bufs[i];
> +				if (outbuf->count == BURST_SIZE)
> +					flush_one_port(outbuf, outp);
> +			}
> +		}
> +	}
> +}
> +
> +
> +static __attribute__((noreturn)) void
> +lcore_worker(struct lcore_params *p)
> +{
> +	struct rte_distributor *d = p->d;
> +	const unsigned id = p->worker_id;
> +	/* for single port, xor_val will be zero so we won't modify the output
> +	 * port, otherwise we send traffic from 0 to 1, 2 to 3, and vice versa
> +	 */
> +	const unsigned xor_val = (rte_eth_dev_count() > 1);
> +	struct rte_mbuf *buf = NULL;
> +
> +	printf("\nCore %u acting as worker core.\n", rte_lcore_id());
> +	for (;;) {
> +		buf = rte_distributor_get_pkt(d, id, buf);
> +		buf->port ^= xor_val;
> +	}
> +}
> +
> +static void
> +int_handler(int sig_num)
> +{
> +	struct rte_eth_stats eth_stats;
> +	unsigned i;
> +
> +	printf("Exiting on signal %d\n", sig_num);
> +
> +	printf("\nRX thread stats:\n");
> +	printf(" - Received:    %"PRIu64"\n", app_stats.rx.rx_pkts);
> +	printf(" - Processed:   %"PRIu64"\n", app_stats.rx.returned_pkts);
> +	printf(" - Enqueued:    %"PRIu64"\n", app_stats.rx.enqueued_pkts);
> +
> +	printf("\nTX thread stats:\n");
> +	printf(" - Dequeued:    %"PRIu64"\n", app_stats.tx.dequeue_pkts);
> +	printf(" - Transmitted: %"PRIu64"\n", app_stats.tx.tx_pkts);
> +
> +	for (i = 0; i < rte_eth_dev_count(); i++) {
> +		rte_eth_stats_get(i, &eth_stats);
> +		printf("\nPort %u stats:\n", i);
> +		printf(" - Pkts in:   %"PRIu64"\n", eth_stats.ipackets);
> +		printf(" - Pkts out:  %"PRIu64"\n", eth_stats.opackets);
> +		printf(" - In Errs:   %"PRIu64"\n", eth_stats.ierrors);
> +		printf(" - Out Errs:  %"PRIu64"\n", eth_stats.oerrors);
> +		printf(" - Mbuf Errs: %"PRIu64"\n", eth_stats.rx_nombuf);
> +	}
> +	exit(0);
rte_exit here?  Also, this is a pretty ungraceful exit strategy as all the
threads you've created and memory you've allocated are just forgotten here.
Given that dpdk mempools are shared, this has the potential to leak lots of
memory if other apps are using the dpdk at the same time that you run this.  You
probably want to use the sigint handler to raise a flag to the tx/rx threads to
shutdown gracefully, and then free your allocated memory and mempool.

Neil